Journal
KST cohorts and the midnight that quietly moves retention
Teams serving users in Korea often mix UTC warehouse timestamps with KST product calendars. A cohort “day” that spans the wrong midnight shifts every retention cell without anyone shipping a feature.
Session Quality Reviews always open with timezone and identity continuity checks. Until those are stable, treating a retention swing as a product win is speculation.
Document the cohort boundary in the same place you document event names. New analysts should not have to reverse-engineer which midnight you meant.
If you changed warehouse export settings recently, rebuild the last two cohorts from raw events before celebrating or mourning the chart.